Indosat Ooredoo Hutchison and Tech Mahindra Collaborate to Develop Garuda, an Advanced LLM for Bahasa Indonesia and Its Variants

  • Indosat Ooredoo Hutchison and Tech Mahindra join forces to develop ‘Garuda,’ a Large Language Model (LLM) for Bahasa Indonesia and its dialects.
  • Garuda aims to preserve Bahasa Indonesia and its dialects, offering personalized customer experiences across various industries.
  • The model will be built on Tech Mahindra’s ‘Project Indus’ principles, enabling conversations in multiple Indic languages and dialects.
  • Garuda will be developed with 16 billion original Bahasa tokens and 1.2 billion parameters, promoting linguistic diversity and inclusivity in the digital realm.
  • Beta version testing will involve Indosat and Bahasa Indonesia speakers, with reinforcement learning techniques enhancing robustness.
  • The partnership signifies a strategic alliance to drive technological innovation, preserve linguistic diversity, and cater to Indonesia’s unique market needs.

Main AI News:

Indosat Ooredoo Hutchison (Indosat or IOH), a prominent digital telecommunications entity in Indonesia, and Tech Mahindra, a frontrunner in digital transformation and business re-engineering solutions, have formally announced a Memorandum of Understanding (MoU) at Mobile World Congress (MWC) 2024. The objective of this collaboration is to construct ‘Garuda,’ a Large Language Model (LLM) aimed at preserving Bahasa Indonesia, the official language of Indonesia, along with its diverse dialects. Garuda’s foundation will draw from Tech Mahindra’s indigenous LLM, ‘Project Indus,’ known for its versatility in various Indic languages and dialects.

The distinct attributes of Garuda are poised to unlock a plethora of applications for Indosat’s clientele across diverse sectors such as healthcare, e-commerce, rural education, banking and finance, agriculture, and telecommunications. Through personalized customer experiences, Garuda will analyze sentiments and behavioral patterns, adapting learning processes based on past interactions. Tech Mahindra will spearhead the collection and curation of Indonesian language data, pre-training it to be deployed as a conversational model for Indosat.

Vikram Sinha, President Director and Chief Executive Officer of Indosat Ooredoo Hutchison, expressed enthusiasm for the ‘Garuda’ initiative, emphasizing its alignment with Indosat Ooredoo Hutchison’s mission of fostering connectivity and empowerment across Indonesia. Sinha highlighted the project’s role in promoting linguistic diversity and bolstering inclusivity in the digital landscape, further underlining their commitment to Indonesia’s digital evolution and the vision of Golden Indonesia 2045.

Garuda’s development will be fueled by 16 billion original Bahasa tokens, contributing to 1.2 billion parameters essential for shaping the model’s comprehension of the Bahasa language. A beta version of Garuda will undergo testing by Indosat and Bahasa Indonesia speakers, followed by refinements utilizing RLHF (Reinforcement Learning from Human Feedback) techniques for enhanced conversational robustness. Additionally, specialized use cases will be addressed using the LIMA (Less is More for Alignment) methodology.

Harshvendra Soin, President – Asia Pacific and Japan Business at Tech Mahindra, noted the burgeoning LLM market is projected to reach 40.8 billion USD by 2029. Soin emphasized the significance of LLMs like Garuda and Indus in facilitating online communication in local dialects, thereby catalyzing digital opportunities. He expressed excitement about the transformative impact Garuda would have on personalized experiences for Indonesians, foreseeing a boost in linguistic diversity and business prospects in the region.

The collaboration between Indosat and Tech Mahindra signifies a strategic partnership aimed at driving technological innovation in Indonesia. This shared vision underscores their commitment to revolutionizing language understanding technology, thereby preserving linguistic diversity and offering tailored communication solutions to meet the unique demands of the Indonesian market.

Conclusion:

The collaboration between Indosat Ooredoo Hutchison and Tech Mahindra to develop the Garuda project not only marks a significant stride towards preserving linguistic diversity but also reflects a strategic alignment to address the unique needs of the Indonesian market. By leveraging advanced language models and technology expertise, the initiative stands poised to revolutionize language understanding, drive personalized experiences, and unlock new business opportunities in Indonesia’s digital landscape.

Source